kookamamie

21 ms·

361.
by kookamamie 3y ago
Trusting the C++ compiler to emit a specific instruction for guaranteed performance - cute, but not realistic.
362.
by kookamamie 3y ago
I more relevant benchmark would be against TensorRT, the current kind of model serving performance on NVIDIA GPUs.
363.
by kookamamie 4y ago
LLAMA is a LLM, very little to do with a model trying to learn MNIST. CNNs in particular gain from using a GPU (or ten), as they're optimizing weights for convolutional/spatial kernels.
364.
by kookamamie 4y ago
> I would guess the constant memory allocation and frees in the training loop are the bottleneck No, the bottleneck would be not utilizing the idling GPU.
365.
by kookamamie 4y ago
> It's a stochastic parrot Stopped reading right there. The author clearly has no clue of how neural networks work or what makes them tick.
366.
by kookamamie 4y ago
It's the font rendering and color scheme. Looks wildly out of date to me.
367.
by kookamamie 4y ago
The font rendering is all over the place. Perhaps it's the screenshots, but just doesn't look good.
368.
by kookamamie 4y ago
They do, but like so many other Rust-projects, they're in their infancy - from the project README: "The project is still in early development, expect bugs, safety issues, and things that don't work".
369.
by kookamamie 4y ago
You missed Zig.
370.
by kookamamie 4y ago
It's quite crappy, to be honest. Even a simple 4x4 block-based BC7 texture compression (1:4 ratio in ARGB) beats this, hands down.
371.
by kookamamie 4y ago
In my estimate, the common joe cannot distinguish between "real" and AI-generated art today. We're only seeing the first generation of AIs right now with the tech such as Midjourney, DALL-E 2, Imagen and Parti. I think it&#x2
372.
by kookamamie 4y ago
You must live in a very special bubble then, given that Android's market share is currently at 72 %.
373.
by kookamamie 4y ago
> CLever Audio Plug-in API A terrible name that'll just look stupid in a couple of years from now.
374.
by kookamamie 4y ago
tcmalloc and mimalloc are the ones I'd be interested seeing comparisons to. The former is the fastest in my case, which is essentially many-core HPC.
375.
by kookamamie 4y ago
Was hoping to read about Screen-Space Reflections.
376.
by kookamamie 5y ago
Agreed. I took notice of OP's narrow view of "tech" and somehow binding web-dev to (assumed here) everything else. Not everything in tech is about gluing frameworks.
377.
by kookamamie 5y ago
Of course you should limit the query duration - just not on some arbitrary function-level like done here. To me this sounds like PHP-level incompetence on API design.
378.
by kookamamie 5y ago
It makes zero sense, if I'm completely honest. Why stop to regexps? What about every other long-running function? This is better handled on a layer above, in a general fashion - imagine a server running some queries and being able to k